Debian Codegraph @

upstream/debian-unstable/man1/perlcall.1

raw
1 .\" -*- mode: troff; coding: utf-8 -*-
2 .\" Automatically generated by Pod::Man 5.0102 (Pod::Simple 3.45)
3 .\"
4 .\" Standard preamble:
5 .\" ========================================================================
6 .de Sp \" Vertical space (when we can't use .PP)
7 .if t .sp .5v
8 .if n .sp
9 ..
10 .de Vb \" Begin verbatim text
11 .ft CW
12 .nf
13 .ne \\$1
14 ..
15 .de Ve \" End verbatim text
16 .ft R
17 .fi
18 ..
19 .\" \*(C` and \*(C' are quotes in nroff, nothing in troff, for use with C<>.
20 .ie n \{\
21 . ds C` ""
22 . ds C' ""
23 'br\}
24 .el\{\
25 . ds C`
26 . ds C'
27 'br\}
28 .\"
29 .\" Escape single quotes in literal strings from groff's Unicode transform.
30 .ie \n(.g .ds Aq \(aq
31 .el .ds Aq '
32 .\"
33 .\" If the F register is >0, we'll generate index entries on stderr for
34 .\" titles (.TH), headers (.SH), subsections (.SS), items (.Ip), and index
35 .\" entries marked with X<> in POD. Of course, you'll have to process the
36 .\" output yourself in some meaningful fashion.
37 .\"
38 .\" Avoid warning from groff about undefined register 'F'.
39 .de IX
40 ..
41 .nr rF 0
42 .if \n(.g .if rF .nr rF 1
43 .if (\n(rF:(\n(.g==0)) \{\
44 . if \nF \{\
45 . de IX
46 . tm Index:\\$1\t\\n%\t"\\$2"
47 ..
48 . if !\nF==2 \{\
49 . nr % 0
50 . nr F 2
51 . \}
52 . \}
53 .\}
54 .rr rF
55 .\" ========================================================================
56 .\"
57 .IX Title "PERLCALL 1"
58 .TH PERLCALL 1 2025-11-16 "perl v5.40.1" "Perl Programmers Reference Guide"
59 .\" For nroff, turn off justification. Always turn off hyphenation; it makes
60 .\" way too many mistakes in technical documents.
61 .if n .ad l
62 .nh
63 .SH NAME
64 perlcall \- Perl calling conventions from C
65 .SH DESCRIPTION
66 .IX Header "DESCRIPTION"
67 The purpose of this document is to show you how to call Perl subroutines
68 directly from C, i.e., how to write \fIcallbacks\fR.
69 .PP
70 Apart from discussing the C interface provided by Perl for writing
71 callbacks the document uses a series of examples to show how the
72 interface actually works in practice. In addition some techniques for
73 coding callbacks are covered.
74 .PP
75 Examples where callbacks are necessary include
76 .IP \(bu 5
77 An Error Handler
78 .Sp
79 You have created an XSUB interface to an application's C API.
80 .Sp
81 A fairly common feature in applications is to allow you to define a C
82 function that will be called whenever something nasty occurs. What we
83 would like is to be able to specify a Perl subroutine that will be
84 called instead.
85 .IP \(bu 5
86 An Event-Driven Program
87 .Sp
88 The classic example of where callbacks are used is when writing an
89 event driven program, such as for an X11 application. In this case
90 you register functions to be called whenever specific events occur,
91 e.g., a mouse button is pressed, the cursor moves into a window or a
92 menu item is selected.
93 .PP
94 Although the techniques described here are applicable when embedding
95 Perl in a C program, this is not the primary goal of this document.
96 There are other details that must be considered and are specific to
97 embedding Perl. For details on embedding Perl in C refer to
98 perlembed.
99 .PP
100 Before you launch yourself head first into the rest of this document,
101 it would be a good idea to have read the following two documents\-\-perlxs
102 and perlguts.
103 .SH "THE CALL_ FUNCTIONS"
104 .IX Header "THE CALL_ FUNCTIONS"
105 Although this stuff is easier to explain using examples, you first need
106 be aware of a few important definitions.
107 .PP
108 Perl has a number of C functions that allow you to call Perl
109 subroutines. They are
110 .PP
111 .Vb 4
112 \& I32 call_sv(SV* sv, I32 flags);
113 \& I32 call_pv(char *subname, I32 flags);
114 \& I32 call_method(char *methname, I32 flags);
115 \& I32 call_argv(char *subname, I32 flags, char **argv);
116 .Ve
117 .PP
118 The key function is \fIcall_sv\fR. All the other functions are
119 fairly simple wrappers which make it easier to call Perl subroutines in
120 special cases. At the end of the day they will all call \fIcall_sv\fR
121 to invoke the Perl subroutine.
122 .PP
123 All the \fIcall_*\fR functions have a \f(CW\*(C`flags\*(C'\fR parameter which is
124 used to pass a bit mask of options to Perl. This bit mask operates
125 identically for each of the functions. The settings available in the
126 bit mask are discussed in "FLAG VALUES".
127 .PP
128 Each of the functions will now be discussed in turn.
129 .IP call_sv 5
130 .IX Item "call_sv"
131 \&\fIcall_sv\fR takes two parameters. The first, \f(CW\*(C`sv\*(C'\fR, is an SV*.
132 This allows you to specify the Perl subroutine to be called either as a
133 C string (which has first been converted to an SV) or a reference to a
134 subroutine. The section, "Using call_sv", shows how you can make
135 use of \fIcall_sv\fR.
136 .IP call_pv 5
137 .IX Item "call_pv"
138 The function, \fIcall_pv\fR, is similar to \fIcall_sv\fR except it
139 expects its first parameter to be a C char* which identifies the Perl
140 subroutine you want to call, e.g., \f(CW\*(C`call_pv("fred", 0)\*(C'\fR. If the
141 subroutine you want to call is in another package, just include the
142 package name in the string, e.g., \f(CW"pkg::fred"\fR.
143 .IP call_method 5
144 .IX Item "call_method"
145 The function \fIcall_method\fR is used to call a method from a Perl
146 class. The parameter \f(CW\*(C`methname\*(C'\fR corresponds to the name of the method
147 to be called. Note that the class that the method belongs to is passed
148 on the Perl stack rather than in the parameter list. This class can be
149 either the name of the class (for a static method) or a reference to an
150 object (for a virtual method). See perlobj for more information on
151 static and virtual methods and "Using call_method" for an example
152 of using \fIcall_method\fR.
153 .IP call_argv 5
154 .IX Item "call_argv"
155 \&\fIcall_argv\fR calls the Perl subroutine specified by the C string
156 stored in the \f(CW\*(C`subname\*(C'\fR parameter. It also takes the usual \f(CW\*(C`flags\*(C'\fR
157 parameter. The final parameter, \f(CW\*(C`argv\*(C'\fR, consists of a NULL-terminated
158 list of C strings to be passed as parameters to the Perl subroutine.
159 See "Using call_argv".
160 .PP
161 All the functions return an integer. This is a count of the number of
162 items returned by the Perl subroutine. The actual items returned by the
163 subroutine are stored on the Perl stack.
164 .PP
165 As a general rule you should \fIalways\fR check the return value from
166 these functions. Even if you are expecting only a particular number of
167 values to be returned from the Perl subroutine, there is nothing to
168 stop someone from doing something unexpected\-\-don't say you haven't
169 been warned.
170 .SH "FLAG VALUES"
171 .IX Header "FLAG VALUES"
172 The \f(CW\*(C`flags\*(C'\fR parameter in all the \fIcall_*\fR functions is one of \f(CW\*(C`G_VOID\*(C'\fR,
173 \&\f(CW\*(C`G_SCALAR\*(C'\fR, or \f(CW\*(C`G_LIST\*(C'\fR, which indicate the call context, OR'ed together
174 with a bit mask of any combination of the other G_* symbols defined below.
175 .SS G_VOID
176 .IX Subsection "G_VOID"
177 Calls the Perl subroutine in a void context.
178 .PP
179 This flag has 2 effects:
180 .IP 1. 5
181 It indicates to the subroutine being called that it is executing in
182 a void context (if it executes \fIwantarray\fR the result will be the
183 undefined value).
184 .IP 2. 5
185 It ensures that nothing is actually returned from the subroutine.
186 .PP
187 The value returned by the \fIcall_*\fR function indicates how many
188 items have been returned by the Perl subroutine\-\-in this case it will
189 be 0.
190 .SS G_SCALAR
191 .IX Subsection "G_SCALAR"
192 Calls the Perl subroutine in a scalar context. This is the default
193 context flag setting for all the \fIcall_*\fR functions.
194 .PP
195 This flag has 2 effects:
196 .IP 1. 5
197 It indicates to the subroutine being called that it is executing in a
198 scalar context (if it executes \fIwantarray\fR the result will be false).
199 .IP 2. 5
200 It ensures that only a scalar is actually returned from the subroutine.
201 The subroutine can, of course, ignore the \fIwantarray\fR and return a
202 list anyway. If so, then only the last element of the list will be
203 returned.
204 .PP
205 The value returned by the \fIcall_*\fR function indicates how many
206 items have been returned by the Perl subroutine \- in this case it will
207 be either 0 or 1.
208 .PP
209 If 0, then you have specified the G_DISCARD flag.
210 .PP
211 If 1, then the item actually returned by the Perl subroutine will be
212 stored on the Perl stack \- the section "Returning a Scalar" shows how
213 to access this value on the stack. Remember that regardless of how
214 many items the Perl subroutine returns, only the last one will be
215 accessible from the stack \- think of the case where only one value is
216 returned as being a list with only one element. Any other items that
217 were returned will not exist by the time control returns from the
218 \&\fIcall_*\fR function. The section "Returning a List in Scalar
219 Context" shows an example of this behavior.
220 .SS G_LIST
221 .IX Subsection "G_LIST"
222 Calls the Perl subroutine in a list context. Prior to Perl version
223 5.35.1 this was called \f(CW\*(C`G_ARRAY\*(C'\fR.
224 .PP
225 As with G_SCALAR, this flag has 2 effects:
226 .IP 1. 5
227 It indicates to the subroutine being called that it is executing in a
228 list context (if it executes \fIwantarray\fR the result will be true).
229 .IP 2. 5
230 It ensures that all items returned from the subroutine will be
231 accessible when control returns from the \fIcall_*\fR function.
232 .PP
233 The value returned by the \fIcall_*\fR function indicates how many
234 items have been returned by the Perl subroutine.
235 .PP
236 If 0, then you have specified the G_DISCARD flag.
237 .PP
238 If not 0, then it will be a count of the number of items returned by
239 the subroutine. These items will be stored on the Perl stack. The
240 section "Returning a List of Values" gives an example of using the
241 G_LIST flag and the mechanics of accessing the returned items from the
242 Perl stack.
243 .SS G_DISCARD
244 .IX Subsection "G_DISCARD"
245 By default, the \fIcall_*\fR functions place the items returned from
246 by the Perl subroutine on the stack. If you are not interested in
247 these items, then setting this flag will make Perl get rid of them
248 automatically for you. Note that it is still possible to indicate a
249 context to the Perl subroutine by using either G_SCALAR or G_LIST.
250 .PP
251 If you do not set this flag then it is \fIvery\fR important that you make
252 sure that any temporaries (i.e., parameters passed to the Perl
253 subroutine and values returned from the subroutine) are disposed of
254 yourself. The section "Returning a Scalar" gives details of how to
255 dispose of these temporaries explicitly and the section "Using Perl to
256 Dispose of Temporaries" discusses the specific circumstances where you
257 can ignore the problem and let Perl deal with it for you.
258 .SS G_NOARGS
259 .IX Subsection "G_NOARGS"
260 Whenever a Perl subroutine is called using one of the \fIcall_*\fR
261 functions, it is assumed by default that parameters are to be passed to
262 the subroutine. If you are not passing any parameters to the Perl
263 subroutine, you can save a bit of time by setting this flag. It has
264 the effect of not creating the \f(CW@_\fR array for the Perl subroutine.
265 .PP
266 Although the functionality provided by this flag may seem
267 straightforward, it should be used only if there is a good reason to do
268 so. The reason for being cautious is that, even if you have specified
269 the G_NOARGS flag, it is still possible for the Perl subroutine that
270 has been called to think that you have passed it parameters.
271 .PP
272 In fact, what can happen is that the Perl subroutine you have called
273 can access the \f(CW@_\fR array from a previous Perl subroutine. This will
274 occur when the code that is executing the \fIcall_*\fR function has
275 itself been called from another Perl subroutine. The code below
276 illustrates this
277 .PP
278 .Vb 2
279 \& sub fred
280 \& { print "@_\en" }
281 \&
282 \& sub joe
283 \& { &fred }
284 \&
285 \& &joe(1,2,3);
286 .Ve
287 .PP
288 This will print
289 .PP
290 .Vb 1
291 \& 1 2 3
292 .Ve
293 .PP
294 What has happened is that \f(CW\*(C`fred\*(C'\fR accesses the \f(CW@_\fR array which
295 belongs to \f(CW\*(C`joe\*(C'\fR.
296 .SS G_EVAL
297 .IX Subsection "G_EVAL"
298 It is possible for the Perl subroutine you are calling to terminate
299 abnormally, e.g., by calling \fIdie\fR explicitly or by not actually
300 existing. By default, when either of these events occurs, the
301 process will terminate immediately. If you want to trap this
302 type of event, specify the G_EVAL flag. It will put an \fIeval { }\fR
303 around the subroutine call.
304 .PP
305 Whenever control returns from the \fIcall_*\fR function you need to
306 check the \f(CW$@\fR variable as you would in a normal Perl script.
307 .PP
308 The value returned from the \fIcall_*\fR function is dependent on
309 what other flags have been specified and whether an error has
310 occurred. Here are all the different cases that can occur:
311 .IP \(bu 5
312 If the \fIcall_*\fR function returns normally, then the value
313 returned is as specified in the previous sections.
314 .IP \(bu 5
315 If G_DISCARD is specified, the return value will always be 0.
316 .IP \(bu 5
317 If G_LIST is specified \fIand\fR an error has occurred, the return value
318 will always be 0.
319 .IP \(bu 5
320 If G_SCALAR is specified \fIand\fR an error has occurred, the return value
321 will be 1 and the value on the top of the stack will be \fIundef\fR. This
322 means that if you have already detected the error by checking \f(CW$@\fR and
323 you want the program to continue, you must remember to pop the \fIundef\fR
324 from the stack.
325 .PP
326 See "Using G_EVAL" for details on using G_EVAL.
327 .SS G_KEEPERR
328 .IX Subsection "G_KEEPERR"
329 Using the G_EVAL flag described above will always set \f(CW$@\fR: clearing
330 it if there was no error, and setting it to describe the error if there
331 was an error in the called code. This is what you want if your intention
332 is to handle possible errors, but sometimes you just want to trap errors
333 and stop them interfering with the rest of the program.
334 .PP
335 This scenario will mostly be applicable to code that is meant to be called
336 from within destructors, asynchronous callbacks, and signal handlers.
337 In such situations, where the code being called has little relation to the
338 surrounding dynamic context, the main program needs to be insulated from
339 errors in the called code, even if they can't be handled intelligently.
340 It may also be useful to do this with code for \f(CW\*(C`_\|_DIE_\|_\*(C'\fR or \f(CW\*(C`_\|_WARN_\|_\*(C'\fR
341 hooks, and \f(CW\*(C`tie\*(C'\fR functions.
342 .PP
343 The G_KEEPERR flag is meant to be used in conjunction with G_EVAL in
344 \&\fIcall_*\fR functions that are used to implement such code, or with
345 \&\f(CW\*(C`eval_sv\*(C'\fR. This flag has no effect on the \f(CW\*(C`call_*\*(C'\fR functions when
346 G_EVAL is not used.
347 .PP
348 When G_KEEPERR is used, any error in the called code will terminate the
349 call as usual, and the error will not propagate beyond the call (as usual
350 for G_EVAL), but it will not go into \f(CW$@\fR. Instead the error will be
351 converted into a warning, prefixed with the string "\et(in cleanup)".
352 This can be disabled using \f(CW\*(C`no warnings \*(Aqmisc\*(Aq\*(C'\fR. If there is no error,
353 \&\f(CW$@\fR will not be cleared.
354 .PP
355 Note that the G_KEEPERR flag does not propagate into inner evals; these
356 may still set \f(CW$@\fR.
357 .PP
358 The G_KEEPERR flag was introduced in Perl version 5.002.
359 .PP
360 See "Using G_KEEPERR" for an example of a situation that warrants the
361 use of this flag.
362 .SS "Determining the Context"
363 .IX Subsection "Determining the Context"
364 As mentioned above, you can determine the context of the currently
365 executing subroutine in Perl with \fIwantarray\fR. The equivalent test
366 can be made in C by using the \f(CW\*(C`GIMME_V\*(C'\fR macro, which returns
367 \&\f(CW\*(C`G_LIST\*(C'\fR if you have been called in a list context, \f(CW\*(C`G_SCALAR\*(C'\fR if
368 in a scalar context, or \f(CW\*(C`G_VOID\*(C'\fR if in a void context (i.e., the
369 return value will not be used). An older version of this macro is
370 called \f(CW\*(C`GIMME\*(C'\fR; in a void context it returns \f(CW\*(C`G_SCALAR\*(C'\fR instead of
371 \&\f(CW\*(C`G_VOID\*(C'\fR. An example of using the \f(CW\*(C`GIMME_V\*(C'\fR macro is shown in
372 section "Using GIMME_V".
373 .SH EXAMPLES
374 .IX Header "EXAMPLES"
375 Enough of the definition talk! Let's have a few examples.
376 .PP
377 Perl provides many macros to assist in accessing the Perl stack.
378 Wherever possible, these macros should always be used when interfacing
379 to Perl internals. We hope this should make the code less vulnerable
380 to any changes made to Perl in the future.
381 .PP
382 Another point worth noting is that in the first series of examples I
383 have made use of only the \fIcall_pv\fR function. This has been done
384 to keep the code simpler and ease you into the topic. Wherever
385 possible, if the choice is between using \fIcall_pv\fR and
386 \&\fIcall_sv\fR, you should always try to use \fIcall_sv\fR. See
387 "Using call_sv" for details.
388 .SS "No Parameters, Nothing Returned"
389 .IX Subsection "No Parameters, Nothing Returned"
390 This first trivial example will call a Perl subroutine, \fIPrintUID\fR, to
391 print out the UID of the process.
392 .PP
393 .Vb 4
394 \& sub PrintUID
395 \& {
396 \& print "UID is $<\en";
397 \& }
398 .Ve
399 .PP
400 and here is a C function to call it
401 .PP
402 .Vb 4
403 \& static void
404 \& call_PrintUID()
405 \& {
406 \& dSP;
407 \&
408 \& PUSHMARK(SP);
409 \& call_pv("PrintUID", G_DISCARD|G_NOARGS);
410 \& }
411 .Ve
412 .PP
413 Simple, eh?
414 .PP
415 A few points to note about this example:
416 .IP 1. 5
417 Ignore \f(CW\*(C`dSP\*(C'\fR and \f(CWPUSHMARK(SP)\fR for now. They will be discussed in
418 the next example.
419 .IP 2. 5
420 We aren't passing any parameters to \fIPrintUID\fR so G_NOARGS can be
421 specified.
422 .IP 3. 5
423 We aren't interested in anything returned from \fIPrintUID\fR, so
424 G_DISCARD is specified. Even if \fIPrintUID\fR was changed to
425 return some value(s), having specified G_DISCARD will mean that they
426 will be wiped by the time control returns from \fIcall_pv\fR.
427 .IP 4. 5
428 As \fIcall_pv\fR is being used, the Perl subroutine is specified as a
429 C string. In this case the subroutine name has been 'hard\-wired' into the
430 code.
431 .IP 5. 5
432 Because we specified G_DISCARD, it is not necessary to check the value
433 returned from \fIcall_pv\fR. It will always be 0.
434 .SS "Passing Parameters"
435 .IX Subsection "Passing Parameters"
436 Now let's make a slightly more complex example. This time we want to
437 call a Perl subroutine, \f(CW\*(C`LeftString\*(C'\fR, which will take 2 parameters\-\-a
438 string ($s) and an integer ($n). The subroutine will simply
439 print the first \f(CW$n\fR characters of the string.
440 .PP
441 So the Perl subroutine would look like this:
442 .PP
443 .Vb 5
444 \& sub LeftString
445 \& {
446 \& my($s, $n) = @_;
447 \& print substr($s, 0, $n), "\en";
448 \& }
449 .Ve
450 .PP
451 The C function required to call \fILeftString\fR would look like this:
452 .PP
453 .Vb 6
454 \& static void
455 \& call_LeftString(a, b)
456 \& char * a;
457 \& int b;
458 \& {
459 \& dSP;
460 \&
461 \& ENTER;
462 \& SAVETMPS;
463 \&
464 \& PUSHMARK(SP);
465 \& EXTEND(SP, 2);
466 \& PUSHs(sv_2mortal(newSVpv(a, 0)));
467 \& PUSHs(sv_2mortal(newSViv(b)));
468 \& PUTBACK;
469 \&
470 \& call_pv("LeftString", G_DISCARD);
471 \&
472 \& FREETMPS;
473 \& LEAVE;
474 \& }
475 .Ve
476 .PP
477 Here are a few notes on the C function \fIcall_LeftString\fR.
478 .IP 1. 5
479 Parameters are passed to the Perl subroutine using the Perl stack.
480 This is the purpose of the code beginning with the line \f(CW\*(C`dSP\*(C'\fR and
481 ending with the line \f(CW\*(C`PUTBACK\*(C'\fR. The \f(CW\*(C`dSP\*(C'\fR declares a local copy
482 of the stack pointer. This local copy should \fBalways\fR be accessed
483 as \f(CW\*(C`SP\*(C'\fR.
484 .IP 2. 5
485 If you are going to put something onto the Perl stack, you need to know
486 where to put it. This is the purpose of the macro \f(CW\*(C`dSP\*(C'\fR\-\-it declares
487 and initializes a \fIlocal\fR copy of the Perl stack pointer.
488 .Sp
489 All the other macros which will be used in this example require you to
490 have used this macro.
491 .Sp
492 The exception to this rule is if you are calling a Perl subroutine
493 directly from an XSUB function. In this case it is not necessary to
494 use the \f(CW\*(C`dSP\*(C'\fR macro explicitly\-\-it will be declared for you
495 automatically.
496 .IP 3. 5
497 Any parameters to be pushed onto the stack should be bracketed by the
498 \&\f(CW\*(C`PUSHMARK\*(C'\fR and \f(CW\*(C`PUTBACK\*(C'\fR macros. The purpose of these two macros, in
499 this context, is to count the number of parameters you are
500 pushing automatically. Then whenever Perl is creating the \f(CW@_\fR array for the
501 subroutine, it knows how big to make it.
502 .Sp
503 The \f(CW\*(C`PUSHMARK\*(C'\fR macro tells Perl to make a mental note of the current
504 stack pointer. Even if you aren't passing any parameters (like the
505 example shown in the section "No Parameters, Nothing Returned") you
506 must still call the \f(CW\*(C`PUSHMARK\*(C'\fR macro before you can call any of the
507 \&\fIcall_*\fR functions\-\-Perl still needs to know that there are no
508 parameters.
509 .Sp
510 The \f(CW\*(C`PUTBACK\*(C'\fR macro sets the global copy of the stack pointer to be
511 the same as our local copy. If we didn't do this, \fIcall_pv\fR
512 wouldn't know where the two parameters we pushed were\-\-remember that
513 up to now all the stack pointer manipulation we have done is with our
514 local copy, \fInot\fR the global copy.
515 .IP 4. 5
516 Next, we come to EXTEND and PUSHs. This is where the parameters
517 actually get pushed onto the stack. In this case we are pushing a
518 string and an integer.
519 .Sp
520 Alternatively you can use the \fBXPUSHs()\fR macro, which combines a
521 \&\f(CW\*(C`EXTEND(SP, 1)\*(C'\fR and \f(CWPUSHs()\fR. This is less efficient if you're
522 pushing multiple values.
523 .Sp
524 See "XSUBs and the Argument Stack" in perlguts for details
525 on how the PUSH macros work.
526 .IP 5. 5
527 Because we created temporary values (by means of \fBsv_2mortal()\fR calls)
528 we will have to tidy up the Perl stack and dispose of mortal SVs.
529 .Sp
530 This is the purpose of
531 .Sp
532 .Vb 2
533 \& ENTER;
534 \& SAVETMPS;
535 .Ve
536 .Sp
537 at the start of the function, and
538 .Sp
539 .Vb 2
540 \& FREETMPS;
541 \& LEAVE;
542 .Ve
543 .Sp
544 at the end. The \f(CW\*(C`ENTER\*(C'\fR/\f(CW\*(C`SAVETMPS\*(C'\fR pair creates a boundary for any
545 temporaries we create. This means that the temporaries we get rid of
546 will be limited to those which were created after these calls.
547 .Sp
548 The \f(CW\*(C`FREETMPS\*(C'\fR/\f(CW\*(C`LEAVE\*(C'\fR pair will get rid of any values returned by
549 the Perl subroutine (see next example), plus it will also dump the
550 mortal SVs we have created. Having \f(CW\*(C`ENTER\*(C'\fR/\f(CW\*(C`SAVETMPS\*(C'\fR at the
551 beginning of the code makes sure that no other mortals are destroyed.
552 .Sp
553 Think of these macros as working a bit like \f(CW\*(C`{\*(C'\fR and \f(CW\*(C`}\*(C'\fR in Perl
554 to limit the scope of local variables.
555 .Sp
556 See the section "Using Perl to Dispose of Temporaries" for details of
557 an alternative to using these macros.
558 .IP 6. 5
559 Finally, \fILeftString\fR can now be called via the \fIcall_pv\fR function.
560 The only flag specified this time is G_DISCARD. Because we are passing
561 2 parameters to the Perl subroutine this time, we have not specified
562 G_NOARGS.
563 .SS "Returning a Scalar"
564 .IX Subsection "Returning a Scalar"
565 Now for an example of dealing with the items returned from a Perl
566 subroutine.
567 .PP
568 Here is a Perl subroutine, \fIAdder\fR, that takes 2 integer parameters
569 and simply returns their sum.
570 .PP
571 .Vb 5
572 \& sub Adder
573 \& {
574 \& my($x, $y) = @_;
575 \& $x + $y;
576 \& }
577 .Ve
578 .PP
579 Because we are now concerned with the return value from \fIAdder\fR, the C
580 function required to call it is now a bit more complex.
581 .PP
582 .Vb 7
583 \& static void
584 \& call_Adder(a, b)
585 \& int a;
586 \& int b;
587 \& {
588 \& dSP;
589 \& int count;
590 \&
591 \& ENTER;
592 \& SAVETMPS;
593 \&
594 \& PUSHMARK(SP);
595 \& EXTEND(SP, 2);
596 \& PUSHs(sv_2mortal(newSViv(a)));
597 \& PUSHs(sv_2mortal(newSViv(b)));
598 \& PUTBACK;
599 \&
600 \& count = call_pv("Adder", G_SCALAR);
601 \&
602 \& SPAGAIN;
603 \&
604 \& if (count != 1)
605 \& croak("Big trouble\en");
606 \&
607 \& printf ("The sum of %d and %d is %d\en", a, b, POPi);
608 \&
609 \& PUTBACK;
610 \& FREETMPS;
611 \& LEAVE;
612 \& }
613 .Ve
614 .PP
615 Points to note this time are
616 .IP 1. 5
617 The only flag specified this time was G_SCALAR. That means that the \f(CW@_\fR
618 array will be created and that the value returned by \fIAdder\fR will
619 still exist after the call to \fIcall_pv\fR.
620 .IP 2. 5
621 The purpose of the macro \f(CW\*(C`SPAGAIN\*(C'\fR is to refresh the local copy of the
622 stack pointer. This is necessary because it is possible that the memory
623 allocated to the Perl stack has been reallocated during the
624 \&\fIcall_pv\fR call.
625 .Sp
626 If you are making use of the Perl stack pointer in your code you must
627 always refresh the local copy using SPAGAIN whenever you make use
628 of the \fIcall_*\fR functions or any other Perl internal function.
629 .IP 3. 5
630 Although only a single value was expected to be returned from \fIAdder\fR,
631 it is still good practice to check the return code from \fIcall_pv\fR
632 anyway.
633 .Sp
634 Expecting a single value is not quite the same as knowing that there
635 will be one. If someone modified \fIAdder\fR to return a list and we
636 didn't check for that possibility and take appropriate action the Perl
637 stack would end up in an inconsistent state. That is something you
638 \&\fIreally\fR don't want to happen ever.
639 .IP 4. 5
640 The \f(CW\*(C`POPi\*(C'\fR macro is used here to pop the return value from the stack.
641 In this case we wanted an integer, so \f(CW\*(C`POPi\*(C'\fR was used.
642 .Sp
643 Here is the complete list of POP macros available, along with the types
644 they return.
645 .Sp
646 .Vb 8
647 \& POPs SV
648 \& POPp pointer (PV)
649 \& POPpbytex pointer to bytes (PV)
650 \& POPn double (NV)
651 \& POPi integer (IV)
652 \& POPu unsigned integer (UV)
653 \& POPl long
654 \& POPul unsigned long
655 .Ve
656 .Sp
657 Since these macros have side-effects don't use them as arguments to
658 macros that may evaluate their argument several times, for example:
659 .Sp
660 .Vb 3
661 \& /* Bad idea, don\*(Aqt do this */
662 \& STRLEN len;
663 \& const char *s = SvPV(POPs, len);
664 .Ve
665 .Sp
666 Instead, use a temporary:
667 .Sp
668 .Vb 3
669 \& STRLEN len;
670 \& SV *sv = POPs;
671 \& const char *s = SvPV(sv, len);
672 .Ve
673 .Sp
674 or a macro that guarantees it will evaluate its arguments only once:
675 .Sp
676 .Vb 2
677 \& STRLEN len;
678 \& const char *s = SvPVx(POPs, len);
679 .Ve
680 .IP 5. 5
681 The final \f(CW\*(C`PUTBACK\*(C'\fR is used to leave the Perl stack in a consistent
682 state before exiting the function. This is necessary because when we
683 popped the return value from the stack with \f(CW\*(C`POPi\*(C'\fR it updated only our
684 local copy of the stack pointer. Remember, \f(CW\*(C`PUTBACK\*(C'\fR sets the global
685 stack pointer to be the same as our local copy.
686 .SS "Returning a List of Values"
687 .IX Subsection "Returning a List of Values"
688 Now, let's extend the previous example to return both the sum of the
689 parameters and the difference.
690 .PP
691 Here is the Perl subroutine
692 .PP
693 .Vb 5
694 \& sub AddSubtract
695 \& {
696 \& my($x, $y) = @_;
697 \& ($x+$y, $x\-$y);
698 \& }
699 .Ve
700 .PP
701 and this is the C function
702 .PP
703 .Vb 7
704 \& static void
705 \& call_AddSubtract(a, b)
706 \& int a;
707 \& int b;
708 \& {
709 \& dSP;
710 \& int count;
711 \&
712 \& ENTER;
713 \& SAVETMPS;
714 \&
715 \& PUSHMARK(SP);
716 \& EXTEND(SP, 2);
717 \& PUSHs(sv_2mortal(newSViv(a)));
718 \& PUSHs(sv_2mortal(newSViv(b)));
719 \& PUTBACK;
720 \&
721 \& count = call_pv("AddSubtract", G_LIST);
722 \&
723 \& SPAGAIN;
724 \&
725 \& if (count != 2)
726 \& croak("Big trouble\en");
727 \&
728 \& printf ("%d \- %d = %d\en", a, b, POPi);
729 \& printf ("%d + %d = %d\en", a, b, POPi);
730 \&
731 \& PUTBACK;
732 \& FREETMPS;
733 \& LEAVE;
734 \& }
735 .Ve
736 .PP
737 If \fIcall_AddSubtract\fR is called like this
738 .PP
739 .Vb 1
740 \& call_AddSubtract(7, 4);
741 .Ve
742 .PP
743 then here is the output
744 .PP
745 .Vb 2
746 \& 7 \- 4 = 3
747 \& 7 + 4 = 11
748 .Ve
749 .PP
750 Notes
751 .IP 1. 5
752 We wanted list context, so G_LIST was used.
753 .IP 2. 5
754 Not surprisingly \f(CW\*(C`POPi\*(C'\fR is used twice this time because we were
755 retrieving 2 values from the stack. The important thing to note is that
756 when using the \f(CW\*(C`POP*\*(C'\fR macros they come off the stack in \fIreverse\fR
757 order.
758 .SS "Returning a List in Scalar Context"
759 .IX Subsection "Returning a List in Scalar Context"
760 Say the Perl subroutine in the previous section was called in a scalar
761 context, like this
762 .PP
763 .Vb 8
764 \& static void
765 \& call_AddSubScalar(a, b)
766 \& int a;
767 \& int b;
768 \& {
769 \& dSP;
770 \& int count;
771 \& int i;
772 \&
773 \& ENTER;
774 \& SAVETMPS;
775 \&
776 \& PUSHMARK(SP);
777 \& EXTEND(SP, 2);
778 \& PUSHs(sv_2mortal(newSViv(a)));
779 \& PUSHs(sv_2mortal(newSViv(b)));
780 \& PUTBACK;
781 \&
782 \& count = call_pv("AddSubtract", G_SCALAR);
783 \&
784 \& SPAGAIN;
785 \&
786 \& printf ("Items Returned = %d\en", count);
787 \&
788 \& for (i = 1; i <= count; ++i)
789 \& printf ("Value %d = %d\en", i, POPi);
790 \&
791 \& PUTBACK;
792 \& FREETMPS;
793 \& LEAVE;
794 \& }
795 .Ve
796 .PP
797 The other modification made is that \fIcall_AddSubScalar\fR will print the
798 number of items returned from the Perl subroutine and their value (for
799 simplicity it assumes that they are integer). So if
800 \&\fIcall_AddSubScalar\fR is called
801 .PP
802 .Vb 1
803 \& call_AddSubScalar(7, 4);
804 .Ve
805 .PP
806 then the output will be
807 .PP
808 .Vb 2
809 \& Items Returned = 1
810 \& Value 1 = 3
811 .Ve
812 .PP
813 In this case the main point to note is that only the last item in the
814 list is returned from the subroutine. \fIAddSubtract\fR actually made it back to
815 \&\fIcall_AddSubScalar\fR.
816 .SS "Returning Data from Perl via the Parameter List"
817 .IX Subsection "Returning Data from Perl via the Parameter List"
818 It is also possible to return values directly via the parameter
819 list\-\-whether it is actually desirable to do it is another matter entirely.
820 .PP
821 The Perl subroutine, \fIInc\fR, below takes 2 parameters and increments
822 each directly.
823 .PP
824 .Vb 5
825 \& sub Inc
826 \& {
827 \& ++ $_[0];
828 \& ++ $_[1];
829 \& }
830 .Ve
831 .PP
832 and here is a C function to call it.
833 .PP
834 .Vb 9
835 \& static void
836 \& call_Inc(a, b)
837 \& int a;
838 \& int b;
839 \& {
840 \& dSP;
841 \& int count;
842 \& SV * sva;
843 \& SV * svb;
844 \&
845 \& ENTER;
846 \& SAVETMPS;
847 \&
848 \& sva = sv_2mortal(newSViv(a));
849 \& svb = sv_2mortal(newSViv(b));
850 \&
851 \& PUSHMARK(SP);
852 \& EXTEND(SP, 2);
853 \& PUSHs(sva);
854 \& PUSHs(svb);
855 \& PUTBACK;
856 \&
857 \& count = call_pv("Inc", G_DISCARD);
858 \&
859 \& if (count != 0)
860 \& croak ("call_Inc: expected 0 values from \*(AqInc\*(Aq, got %d\en",
861 \& count);
862 \&
863 \& printf ("%d + 1 = %d\en", a, SvIV(sva));
864 \& printf ("%d + 1 = %d\en", b, SvIV(svb));
865 \&
866 \& FREETMPS;
867 \& LEAVE;
868 \& }
869 .Ve
870 .PP
871 To be able to access the two parameters that were pushed onto the stack
872 after they return from \fIcall_pv\fR it is necessary to make a note
873 of their addresses\-\-thus the two variables \f(CW\*(C`sva\*(C'\fR and \f(CW\*(C`svb\*(C'\fR.
874 .PP
875 The reason this is necessary is that the area of the Perl stack which
876 held them will very likely have been overwritten by something else by
877 the time control returns from \fIcall_pv\fR.
878 .SS "Using G_EVAL"
879 .IX Subsection "Using G_EVAL"
880 Now an example using G_EVAL. Below is a Perl subroutine which computes
881 the difference of its 2 parameters. If this would result in a negative
882 result, the subroutine calls \fIdie\fR.
883 .PP
884 .Vb 3
885 \& sub Subtract
886 \& {
887 \& my ($x, $y) = @_;
888 \&
889 \& die "death can be fatal\en" if $x < $y;
890 \&
891 \& $x \- $y;
892 \& }
893 .Ve
894 .PP
895 and some C to call it
896 .PP
897 .Vb 8
898 \& static void
899 \& call_Subtract(a, b)
900 \& int a;
901 \& int b;
902 \& {
903 \& dSP;
904 \& int count;
905 \& SV *err_tmp;
906 \&
907 \& ENTER;
908 \& SAVETMPS;
909 \&
910 \& PUSHMARK(SP);
911 \& EXTEND(SP, 2);
912 \& PUSHs(sv_2mortal(newSViv(a)));
913 \& PUSHs(sv_2mortal(newSViv(b)));
914 \& PUTBACK;
915 \&
916 \& count = call_pv("Subtract", G_EVAL|G_SCALAR);
917 \&
918 \& SPAGAIN;
919 \&
920 \& /* Check the eval first */
921 \& err_tmp = ERRSV;
922 \& if (SvTRUE(err_tmp))
923 \& {
924 \& printf ("Uh oh \- %s\en", SvPV_nolen(err_tmp));
925 \& POPs;
926 \& }
927 \& else
928 \& {
929 \& if (count != 1)
930 \& croak("call_Subtract: wanted 1 value from \*(AqSubtract\*(Aq, got %d\en",
931 \& count);
932 \&
933 \& printf ("%d \- %d = %d\en", a, b, POPi);
934 \& }
935 \&
936 \& PUTBACK;
937 \& FREETMPS;
938 \& LEAVE;
939 \& }
940 .Ve
941 .PP
942 If \fIcall_Subtract\fR is called thus
943 .PP
944 .Vb 1
945 \& call_Subtract(4, 5)
946 .Ve
947 .PP
948 the following will be printed
949 .PP
950 .Vb 1
951 \& Uh oh \- death can be fatal
952 .Ve
953 .PP
954 Notes
955 .IP 1. 5
956 We want to be able to catch the \fIdie\fR so we have used the G_EVAL
957 flag. Not specifying this flag would mean that the program would
958 terminate immediately at the \fIdie\fR statement in the subroutine
959 \&\fISubtract\fR.
960 .IP 2. 5
961 The code
962 .Sp
963 .Vb 6
964 \& err_tmp = ERRSV;
965 \& if (SvTRUE(err_tmp))
966 \& {
967 \& printf ("Uh oh \- %s\en", SvPV_nolen(err_tmp));
968 \& POPs;
969 \& }
970 .Ve
971 .Sp
972 is the direct equivalent of this bit of Perl
973 .Sp
974 .Vb 1
975 \& print "Uh oh \- $@\en" if $@;
976 .Ve
977 .Sp
978 \&\f(CW\*(C`PL_errgv\*(C'\fR is a perl global of type \f(CW\*(C`GV *\*(C'\fR that points to the symbol
979 table entry containing the error. \f(CW\*(C`ERRSV\*(C'\fR therefore refers to the C
980 equivalent of \f(CW$@\fR. We use a local temporary, \f(CW\*(C`err_tmp\*(C'\fR, since
981 \&\f(CW\*(C`ERRSV\*(C'\fR is a macro that calls a function, and \f(CWSvTRUE(ERRSV)\fR would
982 end up calling that function multiple times.
983 .IP 3. 5
984 Note that the stack is popped using \f(CW\*(C`POPs\*(C'\fR in the block where
985 \&\f(CWSvTRUE(err_tmp)\fR is true. This is necessary because whenever a
986 \&\fIcall_*\fR function invoked with G_EVAL|G_SCALAR returns an error,
987 the top of the stack holds the value \fIundef\fR. Because we want the
988 program to continue after detecting this error, it is essential that
989 the stack be tidied up by removing the \fIundef\fR.
990 .SS "Using G_KEEPERR"
991 .IX Subsection "Using G_KEEPERR"
992 Consider this rather facetious example, where we have used an XS
993 version of the call_Subtract example above inside a destructor:
994 .PP
995 .Vb 9
996 \& package Foo;
997 \& sub new { bless {}, $_[0] }
998 \& sub Subtract {
999 \& my($x,$y) = @_;
1000 \& die "death can be fatal" if $x < $y;
1001 \& $x \- $y;
1002 \& }
1003 \& sub DESTROY { call_Subtract(5, 4); }
1004 \& sub foo { die "foo dies"; }
1005 \&
1006 \& package main;
1007 \& {
1008 \& my $foo = Foo\->new;
1009 \& eval { $foo\->foo };
1010 \& }
1011 \& print "Saw: $@" if $@; # should be, but isn\*(Aqt
1012 .Ve
1013 .PP
1014 This example will fail to recognize that an error occurred inside the
1015 \&\f(CW\*(C`eval {}\*(C'\fR. Here's why: the call_Subtract code got executed while perl
1016 was cleaning up temporaries when exiting the outer braced block, and because
1017 call_Subtract is implemented with \fIcall_pv\fR using the G_EVAL
1018 flag, it promptly reset \f(CW$@\fR. This results in the failure of the
1019 outermost test for \f(CW$@\fR, and thereby the failure of the error trap.
1020 .PP
1021 Appending the G_KEEPERR flag, so that the \fIcall_pv\fR call in
1022 call_Subtract reads:
1023 .PP
1024 .Vb 1
1025 \& count = call_pv("Subtract", G_EVAL|G_SCALAR|G_KEEPERR);
1026 .Ve
1027 .PP
1028 will preserve the error and restore reliable error handling.
1029 .SS "Using call_sv"
1030 .IX Subsection "Using call_sv"
1031 In all the previous examples I have 'hard\-wired' the name of the Perl
1032 subroutine to be called from C. Most of the time though, it is more
1033 convenient to be able to specify the name of the Perl subroutine from
1034 within the Perl script, and you'll want to use
1035 call_sv.
1036 .PP
1037 Consider the Perl code below
1038 .PP
1039 .Vb 4
1040 \& sub fred
1041 \& {
1042 \& print "Hello there\en";
1043 \& }
1044 \&
1045 \& CallSubPV("fred");
1046 .Ve
1047 .PP
1048 Here is a snippet of XSUB which defines \fICallSubPV\fR.
1049 .PP
1050 .Vb 6
1051 \& void
1052 \& CallSubPV(name)
1053 \& char * name
1054 \& CODE:
1055 \& PUSHMARK(SP);
1056 \& call_pv(name, G_DISCARD|G_NOARGS);
1057 .Ve
1058 .PP
1059 That is fine as far as it goes. The thing is, the Perl subroutine
1060 can be specified as only a string, however, Perl allows references
1061 to subroutines and anonymous subroutines.
1062 This is where \fIcall_sv\fR is useful.
1063 .PP
1064 The code below for \fICallSubSV\fR is identical to \fICallSubPV\fR except
1065 that the \f(CW\*(C`name\*(C'\fR parameter is now defined as an SV* and we use
1066 \&\fIcall_sv\fR instead of \fIcall_pv\fR.
1067 .PP
1068 .Vb 6
1069 \& void
1070 \& CallSubSV(name)
1071 \& SV * name
1072 \& CODE:
1073 \& PUSHMARK(SP);
1074 \& call_sv(name, G_DISCARD|G_NOARGS);
1075 .Ve
1076 .PP
1077 Because we are using an SV to call \fIfred\fR the following can all be used:
1078 .PP
1079 .Vb 5
1080 \& CallSubSV("fred");
1081 \& CallSubSV(\e&fred);
1082 \& $ref = \e&fred;
1083 \& CallSubSV($ref);
1084 \& CallSubSV( sub { print "Hello there\en" } );
1085 .Ve
1086 .PP
1087 As you can see, \fIcall_sv\fR gives you much greater flexibility in
1088 how you can specify the Perl subroutine.
1089 .PP
1090 You should note that, if it is necessary to store the SV (\f(CW\*(C`name\*(C'\fR in the
1091 example above) which corresponds to the Perl subroutine so that it can
1092 be used later in the program, it not enough just to store a copy of the
1093 pointer to the SV. Say the code above had been like this:
1094 .PP
1095 .Vb 1
1096 \& static SV * rememberSub;
1097 \&
1098 \& void
1099 \& SaveSub1(name)
1100 \& SV * name
1101 \& CODE:
1102 \& rememberSub = name;
1103 \&
1104 \& void
1105 \& CallSavedSub1()
1106 \& CODE:
1107 \& PUSHMARK(SP);
1108 \& call_sv(rememberSub, G_DISCARD|G_NOARGS);
1109 .Ve
1110 .PP
1111 The reason this is wrong is that, by the time you come to use the
1112 pointer \f(CW\*(C`rememberSub\*(C'\fR in \f(CW\*(C`CallSavedSub1\*(C'\fR, it may or may not still refer
1113 to the Perl subroutine that was recorded in \f(CW\*(C`SaveSub1\*(C'\fR. This is
1114 particularly true for these cases:
1115 .PP
1116 .Vb 2
1117 \& SaveSub1(\e&fred);
1118 \& CallSavedSub1();
1119 \&
1120 \& SaveSub1( sub { print "Hello there\en" } );
1121 \& CallSavedSub1();
1122 .Ve
1123 .PP
1124 By the time each of the \f(CW\*(C`SaveSub1\*(C'\fR statements above has been executed,
1125 the SV*s which corresponded to the parameters will no longer exist.
1126 Expect an error message from Perl of the form
1127 .PP
1128 .Vb 1
1129 \& Can\*(Aqt use an undefined value as a subroutine reference at ...
1130 .Ve
1131 .PP
1132 for each of the \f(CW\*(C`CallSavedSub1\*(C'\fR lines.
1133 .PP
1134 Similarly, with this code
1135 .PP
1136 .Vb 4
1137 \& $ref = \e&fred;
1138 \& SaveSub1($ref);
1139 \& $ref = 47;
1140 \& CallSavedSub1();
1141 .Ve
1142 .PP
1143 you can expect one of these messages (which you actually get is dependent on
1144 the version of Perl you are using)
1145 .PP
1146 .Vb 2
1147 \& Not a CODE reference at ...
1148 \& Undefined subroutine &main::47 called ...
1149 .Ve
1150 .PP
1151 The variable \f(CW$ref\fR may have referred to the subroutine \f(CW\*(C`fred\*(C'\fR
1152 whenever the call to \f(CW\*(C`SaveSub1\*(C'\fR was made but by the time
1153 \&\f(CW\*(C`CallSavedSub1\*(C'\fR gets called it now holds the number \f(CW47\fR. Because we
1154 saved only a pointer to the original SV in \f(CW\*(C`SaveSub1\*(C'\fR, any changes to
1155 \&\f(CW$ref\fR will be tracked by the pointer \f(CW\*(C`rememberSub\*(C'\fR. This means that
1156 whenever \f(CW\*(C`CallSavedSub1\*(C'\fR gets called, it will attempt to execute the
1157 code which is referenced by the SV* \f(CW\*(C`rememberSub\*(C'\fR. In this case
1158 though, it now refers to the integer \f(CW47\fR, so expect Perl to complain
1159 loudly.
1160 .PP
1161 A similar but more subtle problem is illustrated with this code:
1162 .PP
1163 .Vb 4
1164 \& $ref = \e&fred;
1165 \& SaveSub1($ref);
1166 \& $ref = \e&joe;
1167 \& CallSavedSub1();
1168 .Ve
1169 .PP
1170 This time whenever \f(CW\*(C`CallSavedSub1\*(C'\fR gets called it will execute the Perl
1171 subroutine \f(CW\*(C`joe\*(C'\fR (assuming it exists) rather than \f(CW\*(C`fred\*(C'\fR as was
1172 originally requested in the call to \f(CW\*(C`SaveSub1\*(C'\fR.
1173 .PP
1174 To get around these problems it is necessary to take a full copy of the
1175 SV. The code below shows \f(CW\*(C`SaveSub2\*(C'\fR modified to do that.
1176 .PP
1177 .Vb 2
1178 \& /* this isn\*(Aqt thread\-safe */
1179 \& static SV * keepSub = (SV*)NULL;
1180 \&
1181 \& void
1182 \& SaveSub2(name)
1183 \& SV * name
1184 \& CODE:
1185 \& /* Take a copy of the callback */
1186 \& if (keepSub == (SV*)NULL)
1187 \& /* First time, so create a new SV */
1188 \& keepSub = newSVsv(name);
1189 \& else
1190 \& /* Been here before, so overwrite */
1191 \& SvSetSV(keepSub, name);
1192 \&
1193 \& void
1194 \& CallSavedSub2()
1195 \& CODE:
1196 \& PUSHMARK(SP);
1197 \& call_sv(keepSub, G_DISCARD|G_NOARGS);
1198 .Ve
1199 .PP
1200 To avoid creating a new SV every time \f(CW\*(C`SaveSub2\*(C'\fR is called,
1201 the function first checks to see if it has been called before. If not,
1202 then space for a new SV is allocated and the reference to the Perl
1203 subroutine \f(CW\*(C`name\*(C'\fR is copied to the variable \f(CW\*(C`keepSub\*(C'\fR in one
1204 operation using \f(CW\*(C`newSVsv\*(C'\fR. Thereafter, whenever \f(CW\*(C`SaveSub2\*(C'\fR is called,
1205 the existing SV, \f(CW\*(C`keepSub\*(C'\fR, is overwritten with the new value using
1206 \&\f(CW\*(C`SvSetSV\*(C'\fR.
1207 .PP
1208 Note: using a static or global variable to store the SV isn't
1209 thread-safe. You can either use the \f(CW\*(C`MY_CXT\*(C'\fR mechanism documented in
1210 "Safely Storing Static Data in XS" in perlxs which is fast, or store the
1211 values in perl global variables, using \fBget_sv()\fR, which is much slower.
1212 .SS "Using call_argv"
1213 .IX Subsection "Using call_argv"
1214 Here is a Perl subroutine which prints whatever parameters are passed
1215 to it.
1216 .PP
1217 .Vb 3
1218 \& sub PrintList
1219 \& {
1220 \& my(@list) = @_;
1221 \&
1222 \& foreach (@list) { print "$_\en" }
1223 \& }
1224 .Ve
1225 .PP
1226 And here is an example of \fIcall_argv\fR which will call
1227 \&\fIPrintList\fR.
1228 .PP
1229 .Vb 1
1230 \& static char * words[] = {"alpha", "beta", "gamma", "delta", NULL};
1231 \&
1232 \& static void
1233 \& call_PrintList()
1234 \& {
1235 \& call_argv("PrintList", G_DISCARD, words);
1236 \& }
1237 .Ve
1238 .PP
1239 Note that it is not necessary to call \f(CW\*(C`PUSHMARK\*(C'\fR in this instance.
1240 This is because \fIcall_argv\fR will do it for you.
1241 .SS "Using call_method"
1242 .IX Subsection "Using call_method"
1243 Consider the following Perl code:
1244 .PP
1245 .Vb 2
1246 \& {
1247 \& package Mine;
1248 \&
1249 \& sub new
1250 \& {
1251 \& my($type) = shift;
1252 \& bless [@_]
1253 \& }
1254 \&
1255 \& sub Display
1256 \& {
1257 \& my ($self, $index) = @_;
1258 \& print "$index: $$self[$index]\en";
1259 \& }
1260 \&
1261 \& sub PrintID
1262 \& {
1263 \& my($class) = @_;
1264 \& print "This is Class $class version 1.0\en";
1265 \& }
1266 \& }
1267 .Ve
1268 .PP
1269 It implements just a very simple class to manage an array. Apart from
1270 the constructor, \f(CW\*(C`new\*(C'\fR, it declares methods, one static and one
1271 virtual. The static method, \f(CW\*(C`PrintID\*(C'\fR, prints out simply the class
1272 name and a version number. The virtual method, \f(CW\*(C`Display\*(C'\fR, prints out a
1273 single element of the array. Here is an all-Perl example of using it.
1274 .PP
1275 .Vb 3
1276 \& $x = Mine\->new(\*(Aqred\*(Aq, \*(Aqgreen\*(Aq, \*(Aqblue\*(Aq);
1277 \& $x\->Display(1);
1278 \& Mine\->PrintID;
1279 .Ve
1280 .PP
1281 will print
1282 .PP
1283 .Vb 2
1284 \& 1: green
1285 \& This is Class Mine version 1.0
1286 .Ve
1287 .PP
1288 Calling a Perl method from C is fairly straightforward. The following
1289 things are required:
1290 .IP \(bu 5
1291 A reference to the object for a virtual method or the name of the class
1292 for a static method
1293 .IP \(bu 5
1294 The name of the method
1295 .IP \(bu 5
1296 Any other parameters specific to the method
1297 .PP
1298 Here is a simple XSUB which illustrates the mechanics of calling both
1299 the \f(CW\*(C`PrintID\*(C'\fR and \f(CW\*(C`Display\*(C'\fR methods from C.
1300 .PP
1301 .Vb 11
1302 \& void
1303 \& call_Method(ref, method, index)
1304 \& SV * ref
1305 \& char * method
1306 \& int index
1307 \& CODE:
1308 \& PUSHMARK(SP);
1309 \& EXTEND(SP, 2);
1310 \& PUSHs(ref);
1311 \& PUSHs(sv_2mortal(newSViv(index)));
1312 \& PUTBACK;
1313 \&
1314 \& call_method(method, G_DISCARD);
1315 \&
1316 \& void
1317 \& call_PrintID(class, method)
1318 \& char * class
1319 \& char * method
1320 \& CODE:
1321 \& PUSHMARK(SP);
1322 \& XPUSHs(sv_2mortal(newSVpv(class, 0)));
1323 \& PUTBACK;
1324 \&
1325 \& call_method(method, G_DISCARD);
1326 .Ve
1327 .PP
1328 So the methods \f(CW\*(C`PrintID\*(C'\fR and \f(CW\*(C`Display\*(C'\fR can be invoked like this:
1329 .PP
1330 .Vb 3
1331 \& $x = Mine\->new(\*(Aqred\*(Aq, \*(Aqgreen\*(Aq, \*(Aqblue\*(Aq);
1332 \& call_Method($x, \*(AqDisplay\*(Aq, 1);
1333 \& call_PrintID(\*(AqMine\*(Aq, \*(AqPrintID\*(Aq);
1334 .Ve
1335 .PP
1336 The only thing to note is that, in both the static and virtual methods,
1337 the method name is not passed via the stack\-\-it is used as the first
1338 parameter to \fIcall_method\fR.
1339 .SS "Using GIMME_V"
1340 .IX Subsection "Using GIMME_V"
1341 Here is a trivial XSUB which prints the context in which it is
1342 currently executing.
1343 .PP
1344 .Vb 10
1345 \& void
1346 \& PrintContext()
1347 \& CODE:
1348 \& U8 gimme = GIMME_V;
1349 \& if (gimme == G_VOID)
1350 \& printf ("Context is Void\en");
1351 \& else if (gimme == G_SCALAR)
1352 \& printf ("Context is Scalar\en");
1353 \& else
1354 \& printf ("Context is Array\en");
1355 .Ve
1356 .PP
1357 And here is some Perl to test it.
1358 .PP
1359 .Vb 3
1360 \& PrintContext;
1361 \& $x = PrintContext;
1362 \& @a = PrintContext;
1363 .Ve
1364 .PP
1365 The output from that will be
1366 .PP
1367 .Vb 3
1368 \& Context is Void
1369 \& Context is Scalar
1370 \& Context is Array
1371 .Ve
1372 .SS "Using Perl to Dispose of Temporaries"
1373 .IX Subsection "Using Perl to Dispose of Temporaries"
1374 In the examples given to date, any temporaries created in the callback
1375 (i.e., parameters passed on the stack to the \fIcall_*\fR function or
1376 values returned via the stack) have been freed by one of these methods:
1377 .IP \(bu 5
1378 Specifying the G_DISCARD flag with \fIcall_*\fR
1379 .IP \(bu 5
1380 Explicitly using the \f(CW\*(C`ENTER\*(C'\fR/\f(CW\*(C`SAVETMPS\*(C'\fR\-\-\f(CW\*(C`FREETMPS\*(C'\fR/\f(CW\*(C`LEAVE\*(C'\fR pairing
1381 .PP
1382 There is another method which can be used, namely letting Perl do it
1383 for you automatically whenever it regains control after the callback
1384 has terminated. This is done by simply not using the
1385 .PP
1386 .Vb 5
1387 \& ENTER;
1388 \& SAVETMPS;
1389 \& ...
1390 \& FREETMPS;
1391 \& LEAVE;
1392 .Ve
1393 .PP
1394 sequence in the callback (and not, of course, specifying the G_DISCARD
1395 flag).
1396 .PP
1397 If you are going to use this method you have to be aware of a possible
1398 memory leak which can arise under very specific circumstances. To
1399 explain these circumstances you need to know a bit about the flow of
1400 control between Perl and the callback routine.
1401 .PP
1402 The examples given at the start of the document (an error handler and
1403 an event driven program) are typical of the two main sorts of flow
1404 control that you are likely to encounter with callbacks. There is a
1405 very important distinction between them, so pay attention.
1406 .PP
1407 In the first example, an error handler, the flow of control could be as
1408 follows. You have created an interface to an external library.
1409 Control can reach the external library like this
1410 .PP
1411 .Vb 1
1412 \& perl \-\-> XSUB \-\-> external library
1413 .Ve
1414 .PP
1415 Whilst control is in the library, an error condition occurs. You have
1416 previously set up a Perl callback to handle this situation, so it will
1417 get executed. Once the callback has finished, control will drop back to
1418 Perl again. Here is what the flow of control will be like in that
1419 situation
1420 .PP
1421 .Vb 7
1422 \& perl \-\-> XSUB \-\-> external library
1423 \& ...
1424 \& error occurs
1425 \& ...
1426 \& external library \-\-> call_* \-\-> perl
1427 \& |
1428 \& perl <\-\- XSUB <\-\- external library <\-\- call_* <\-\-\-\-+
1429 .Ve
1430 .PP
1431 After processing of the error using \fIcall_*\fR is completed,
1432 control reverts back to Perl more or less immediately.
1433 .PP
1434 In the diagram, the further right you go the more deeply nested the
1435 scope is. It is only when control is back with perl on the extreme
1436 left of the diagram that you will have dropped back to the enclosing
1437 scope and any temporaries you have left hanging around will be freed.
1438 .PP
1439 In the second example, an event driven program, the flow of control
1440 will be more like this
1441 .PP
1442 .Vb 10
1443 \& perl \-\-> XSUB \-\-> event handler
1444 \& ...
1445 \& event handler \-\-> call_* \-\-> perl
1446 \& |
1447 \& event handler <\-\- call_* <\-\-\-\-+
1448 \& ...
1449 \& event handler \-\-> call_* \-\-> perl
1450 \& |
1451 \& event handler <\-\- call_* <\-\-\-\-+
1452 \& ...
1453 \& event handler \-\-> call_* \-\-> perl
1454 \& |
1455 \& event handler <\-\- call_* <\-\-\-\-+
1456 .Ve
1457 .PP
1458 In this case the flow of control can consist of only the repeated
1459 sequence
1460 .PP
1461 .Vb 1
1462 \& event handler \-\-> call_* \-\-> perl
1463 .Ve
1464 .PP
1465 for practically the complete duration of the program. This means that
1466 control may \fInever\fR drop back to the surrounding scope in Perl at the
1467 extreme left.
1468 .PP
1469 So what is the big problem? Well, if you are expecting Perl to tidy up
1470 those temporaries for you, you might be in for a long wait. For Perl
1471 to dispose of your temporaries, control must drop back to the
1472 enclosing scope at some stage. In the event driven scenario that may
1473 never happen. This means that, as time goes on, your program will
1474 create more and more temporaries, none of which will ever be freed. As
1475 each of these temporaries consumes some memory your program will
1476 eventually consume all the available memory in your system\-\-kapow!
1477 .PP
1478 So here is the bottom line\-\-if you are sure that control will revert
1479 back to the enclosing Perl scope fairly quickly after the end of your
1480 callback, then it isn't absolutely necessary to dispose explicitly of
1481 any temporaries you may have created. Mind you, if you are at all
1482 uncertain about what to do, it doesn't do any harm to tidy up anyway.
1483 .SS "Strategies for Storing Callback Context Information"
1484 .IX Subsection "Strategies for Storing Callback Context Information"
1485 Potentially one of the trickiest problems to overcome when designing a
1486 callback interface can be figuring out how to store the mapping between
1487 the C callback function and the Perl equivalent.
1488 .PP
1489 To help understand why this can be a real problem first consider how a
1490 callback is set up in an all C environment. Typically a C API will
1491 provide a function to register a callback. This will expect a pointer
1492 to a function as one of its parameters. Below is a call to a
1493 hypothetical function \f(CW\*(C`register_fatal\*(C'\fR which registers the C function
1494 to get called when a fatal error occurs.
1495 .PP
1496 .Vb 1
1497 \& register_fatal(cb1);
1498 .Ve
1499 .PP
1500 The single parameter \f(CW\*(C`cb1\*(C'\fR is a pointer to a function, so you must
1501 have defined \f(CW\*(C`cb1\*(C'\fR in your code, say something like this
1502 .PP
1503 .Vb 6
1504 \& static void
1505 \& cb1()
1506 \& {
1507 \& printf ("Fatal Error\en");
1508 \& exit(1);
1509 \& }
1510 .Ve
1511 .PP
1512 Now change that to call a Perl subroutine instead
1513 .PP
1514 .Vb 1
1515 \& static SV * callback = (SV*)NULL;
1516 \&
1517 \& static void
1518 \& cb1()
1519 \& {
1520 \& dSP;
1521 \&
1522 \& PUSHMARK(SP);
1523 \&
1524 \& /* Call the Perl sub to process the callback */
1525 \& call_sv(callback, G_DISCARD);
1526 \& }
1527 \&
1528 \&
1529 \& void
1530 \& register_fatal(fn)
1531 \& SV * fn
1532 \& CODE:
1533 \& /* Remember the Perl sub */
1534 \& if (callback == (SV*)NULL)
1535 \& callback = newSVsv(fn);
1536 \& else
1537 \& SvSetSV(callback, fn);
1538 \&
1539 \& /* register the callback with the external library */
1540 \& register_fatal(cb1);
1541 .Ve
1542 .PP
1543 where the Perl equivalent of \f(CW\*(C`register_fatal\*(C'\fR and the callback it
1544 registers, \f(CW\*(C`pcb1\*(C'\fR, might look like this
1545 .PP
1546 .Vb 2
1547 \& # Register the sub pcb1
1548 \& register_fatal(\e&pcb1);
1549 \&
1550 \& sub pcb1
1551 \& {
1552 \& die "I\*(Aqm dying...\en";
1553 \& }
1554 .Ve
1555 .PP
1556 The mapping between the C callback and the Perl equivalent is stored in
1557 the global variable \f(CW\*(C`callback\*(C'\fR.
1558 .PP
1559 This will be adequate if you ever need to have only one callback
1560 registered at any time. An example could be an error handler like the
1561 code sketched out above. Remember though, repeated calls to
1562 \&\f(CW\*(C`register_fatal\*(C'\fR will replace the previously registered callback
1563 function with the new one.
1564 .PP
1565 Say for example you want to interface to a library which allows asynchronous
1566 file i/o. In this case you may be able to register a callback whenever
1567 a read operation has completed. To be of any use we want to be able to
1568 call separate Perl subroutines for each file that is opened. As it
1569 stands, the error handler example above would not be adequate as it
1570 allows only a single callback to be defined at any time. What we
1571 require is a means of storing the mapping between the opened file and
1572 the Perl subroutine we want to be called for that file.
1573 .PP
1574 Say the i/o library has a function \f(CW\*(C`asynch_read\*(C'\fR which associates a C
1575 function \f(CW\*(C`ProcessRead\*(C'\fR with a file handle \f(CW\*(C`fh\*(C'\fR\-\-this assumes that it
1576 has also provided some routine to open the file and so obtain the file
1577 handle.
1578 .PP
1579 .Vb 1
1580 \& asynch_read(fh, ProcessRead)
1581 .Ve
1582 .PP
1583 This may expect the C \fIProcessRead\fR function of this form
1584 .PP
1585 .Vb 7
1586 \& void
1587 \& ProcessRead(fh, buffer)
1588 \& int fh;
1589 \& char * buffer;
1590 \& {
1591 \& ...
1592 \& }
1593 .Ve
1594 .PP
1595 To provide a Perl interface to this library we need to be able to map
1596 between the \f(CW\*(C`fh\*(C'\fR parameter and the Perl subroutine we want called. A
1597 hash is a convenient mechanism for storing this mapping. The code
1598 below shows a possible implementation
1599 .PP
1600 .Vb 1
1601 \& static HV * Mapping = (HV*)NULL;
1602 \&
1603 \& void
1604 \& asynch_read(fh, callback)
1605 \& int fh
1606 \& SV * callback
1607 \& CODE:
1608 \& /* If the hash doesn\*(Aqt already exist, create it */
1609 \& if (Mapping == (HV*)NULL)
1610 \& Mapping = newHV();
1611 \&
1612 \& /* Save the fh \-> callback mapping */
1613 \& hv_store(Mapping, (char*)&fh, sizeof(fh), newSVsv(callback), 0);
1614 \&
1615 \& /* Register with the C Library */
1616 \& asynch_read(fh, asynch_read_if);
1617 .Ve
1618 .PP
1619 and \f(CW\*(C`asynch_read_if\*(C'\fR could look like this
1620 .PP
1621 .Vb 7
1622 \& static void
1623 \& asynch_read_if(fh, buffer)
1624 \& int fh;
1625 \& char * buffer;
1626 \& {
1627 \& dSP;
1628 \& SV ** sv;
1629 \&
1630 \& /* Get the callback associated with fh */
1631 \& sv = hv_fetch(Mapping, (char*)&fh , sizeof(fh), FALSE);
1632 \& if (sv == (SV**)NULL)
1633 \& croak("Internal error...\en");
1634 \&
1635 \& PUSHMARK(SP);
1636 \& EXTEND(SP, 2);
1637 \& PUSHs(sv_2mortal(newSViv(fh)));
1638 \& PUSHs(sv_2mortal(newSVpv(buffer, 0)));
1639 \& PUTBACK;
1640 \&
1641 \& /* Call the Perl sub */
1642 \& call_sv(*sv, G_DISCARD);
1643 \& }
1644 .Ve
1645 .PP
1646 For completeness, here is \f(CW\*(C`asynch_close\*(C'\fR. This shows how to remove
1647 the entry from the hash \f(CW\*(C`Mapping\*(C'\fR.
1648 .PP
1649 .Vb 6
1650 \& void
1651 \& asynch_close(fh)
1652 \& int fh
1653 \& CODE:
1654 \& /* Remove the entry from the hash */
1655 \& (void) hv_delete(Mapping, (char*)&fh, sizeof(fh), G_DISCARD);
1656 \&
1657 \& /* Now call the real asynch_close */
1658 \& asynch_close(fh);
1659 .Ve
1660 .PP
1661 So the Perl interface would look like this
1662 .PP
1663 .Vb 4
1664 \& sub callback1
1665 \& {
1666 \& my($handle, $buffer) = @_;
1667 \& }
1668 \&
1669 \& # Register the Perl callback
1670 \& asynch_read($fh, \e&callback1);
1671 \&
1672 \& asynch_close($fh);
1673 .Ve
1674 .PP
1675 The mapping between the C callback and Perl is stored in the global
1676 hash \f(CW\*(C`Mapping\*(C'\fR this time. Using a hash has the distinct advantage that
1677 it allows an unlimited number of callbacks to be registered.
1678 .PP
1679 What if the interface provided by the C callback doesn't contain a
1680 parameter which allows the file handle to Perl subroutine mapping? Say
1681 in the asynchronous i/o package, the callback function gets passed only
1682 the \f(CW\*(C`buffer\*(C'\fR parameter like this
1683 .PP
1684 .Vb 6
1685 \& void
1686 \& ProcessRead(buffer)
1687 \& char * buffer;
1688 \& {
1689 \& ...
1690 \& }
1691 .Ve
1692 .PP
1693 Without the file handle there is no straightforward way to map from the
1694 C callback to the Perl subroutine.
1695 .PP
1696 In this case a possible way around this problem is to predefine a
1697 series of C functions to act as the interface to Perl, thus
1698 .PP
1699 .Vb 3
1700 \& #define MAX_CB 3
1701 \& #define NULL_HANDLE \-1
1702 \& typedef void (*FnMap)();
1703 \&
1704 \& struct MapStruct {
1705 \& FnMap Function;
1706 \& SV * PerlSub;
1707 \& int Handle;
1708 \& };
1709 \&
1710 \& static void fn1();
1711 \& static void fn2();
1712 \& static void fn3();
1713 \&
1714 \& static struct MapStruct Map [MAX_CB] =
1715 \& {
1716 \& { fn1, NULL, NULL_HANDLE },
1717 \& { fn2, NULL, NULL_HANDLE },
1718 \& { fn3, NULL, NULL_HANDLE }
1719 \& };
1720 \&
1721 \& static void
1722 \& Pcb(index, buffer)
1723 \& int index;
1724 \& char * buffer;
1725 \& {
1726 \& dSP;
1727 \&
1728 \& PUSHMARK(SP);
1729 \& XPUSHs(sv_2mortal(newSVpv(buffer, 0)));
1730 \& PUTBACK;
1731 \&
1732 \& /* Call the Perl sub */
1733 \& call_sv(Map[index].PerlSub, G_DISCARD);
1734 \& }
1735 \&
1736 \& static void
1737 \& fn1(buffer)
1738 \& char * buffer;
1739 \& {
1740 \& Pcb(0, buffer);
1741 \& }
1742 \&
1743 \& static void
1744 \& fn2(buffer)
1745 \& char * buffer;
1746 \& {
1747 \& Pcb(1, buffer);
1748 \& }
1749 \&
1750 \& static void
1751 \& fn3(buffer)
1752 \& char * buffer;
1753 \& {
1754 \& Pcb(2, buffer);
1755 \& }
1756 \&
1757 \& void
1758 \& array_asynch_read(fh, callback)
1759 \& int fh
1760 \& SV * callback
1761 \& CODE:
1762 \& int index;
1763 \& int null_index = MAX_CB;
1764 \&
1765 \& /* Find the same handle or an empty entry */
1766 \& for (index = 0; index < MAX_CB; ++index)
1767 \& {
1768 \& if (Map[index].Handle == fh)
1769 \& break;
1770 \&
1771 \& if (Map[index].Handle == NULL_HANDLE)
1772 \& null_index = index;
1773 \& }
1774 \&
1775 \& if (index == MAX_CB && null_index == MAX_CB)
1776 \& croak ("Too many callback functions registered\en");
1777 \&
1778 \& if (index == MAX_CB)
1779 \& index = null_index;
1780 \&
1781 \& /* Save the file handle */
1782 \& Map[index].Handle = fh;
1783 \&
1784 \& /* Remember the Perl sub */
1785 \& if (Map[index].PerlSub == (SV*)NULL)
1786 \& Map[index].PerlSub = newSVsv(callback);
1787 \& else
1788 \& SvSetSV(Map[index].PerlSub, callback);
1789 \&
1790 \& asynch_read(fh, Map[index].Function);
1791 \&
1792 \& void
1793 \& array_asynch_close(fh)
1794 \& int fh
1795 \& CODE:
1796 \& int index;
1797 \&
1798 \& /* Find the file handle */
1799 \& for (index = 0; index < MAX_CB; ++ index)
1800 \& if (Map[index].Handle == fh)
1801 \& break;
1802 \&
1803 \& if (index == MAX_CB)
1804 \& croak ("could not close fh %d\en", fh);
1805 \&
1806 \& Map[index].Handle = NULL_HANDLE;
1807 \& SvREFCNT_dec(Map[index].PerlSub);
1808 \& Map[index].PerlSub = (SV*)NULL;
1809 \&
1810 \& asynch_close(fh);
1811 .Ve
1812 .PP
1813 In this case the functions \f(CW\*(C`fn1\*(C'\fR, \f(CW\*(C`fn2\*(C'\fR, and \f(CW\*(C`fn3\*(C'\fR are used to
1814 remember the Perl subroutine to be called. Each of the functions holds
1815 a separate hard-wired index which is used in the function \f(CW\*(C`Pcb\*(C'\fR to
1816 access the \f(CW\*(C`Map\*(C'\fR array and actually call the Perl subroutine.
1817 .PP
1818 There are some obvious disadvantages with this technique.
1819 .PP
1820 Firstly, the code is considerably more complex than with the previous
1821 example.
1822 .PP
1823 Secondly, there is a hard-wired limit (in this case 3) to the number of
1824 callbacks that can exist simultaneously. The only way to increase the
1825 limit is by modifying the code to add more functions and then
1826 recompiling. None the less, as long as the number of functions is
1827 chosen with some care, it is still a workable solution and in some
1828 cases is the only one available.
1829 .PP
1830 To summarize, here are a number of possible methods for you to consider
1831 for storing the mapping between C and the Perl callback
1832 .IP "1. Ignore the problem \- Allow only 1 callback" 5
1833 .IX Item "1. Ignore the problem - Allow only 1 callback"
1834 For a lot of situations, like interfacing to an error handler, this may
1835 be a perfectly adequate solution.
1836 .IP "2. Create a sequence of callbacks \- hard wired limit" 5
1837 .IX Item "2. Create a sequence of callbacks - hard wired limit"
1838 If it is impossible to tell from the parameters passed back from the C
1839 callback what the context is, then you may need to create a sequence of C
1840 callback interface functions, and store pointers to each in an array.
1841 .IP "3. Use a parameter to map to the Perl callback" 5
1842 .IX Item "3. Use a parameter to map to the Perl callback"
1843 A hash is an ideal mechanism to store the mapping between C and Perl.
1844 .SS "Alternate Stack Manipulation"
1845 .IX Subsection "Alternate Stack Manipulation"
1846 Although I have made use of only the \f(CW\*(C`POP*\*(C'\fR macros to access values
1847 returned from Perl subroutines, it is also possible to bypass these
1848 macros and read the stack using the \f(CW\*(C`ST\*(C'\fR macro (See perlxs for a
1849 full description of the \f(CW\*(C`ST\*(C'\fR macro).
1850 .PP
1851 Most of the time the \f(CW\*(C`POP*\*(C'\fR macros should be adequate; the main
1852 problem with them is that they force you to process the returned values
1853 in sequence. This may not be the most suitable way to process the
1854 values in some cases. What we want is to be able to access the stack in
1855 a random order. The \f(CW\*(C`ST\*(C'\fR macro as used when coding an XSUB is ideal
1856 for this purpose.
1857 .PP
1858 The code below is the example given in the section "Returning a List
1859 of Values" recoded to use \f(CW\*(C`ST\*(C'\fR instead of \f(CW\*(C`POP*\*(C'\fR.
1860 .PP
1861 .Vb 8
1862 \& static void
1863 \& call_AddSubtract2(a, b)
1864 \& int a;
1865 \& int b;
1866 \& {
1867 \& dSP;
1868 \& I32 ax;
1869 \& int count;
1870 \&
1871 \& ENTER;
1872 \& SAVETMPS;
1873 \&
1874 \& PUSHMARK(SP);
1875 \& EXTEND(SP, 2);
1876 \& PUSHs(sv_2mortal(newSViv(a)));
1877 \& PUSHs(sv_2mortal(newSViv(b)));
1878 \& PUTBACK;
1879 \&
1880 \& count = call_pv("AddSubtract", G_LIST);
1881 \&
1882 \& SPAGAIN;
1883 \& SP \-= count;
1884 \& ax = (SP \- PL_stack_base) + 1;
1885 \&
1886 \& if (count != 2)
1887 \& croak("Big trouble\en");
1888 \&
1889 \& printf ("%d + %d = %d\en", a, b, SvIV(ST(0)));
1890 \& printf ("%d \- %d = %d\en", a, b, SvIV(ST(1)));
1891 \&
1892 \& PUTBACK;
1893 \& FREETMPS;
1894 \& LEAVE;
1895 \& }
1896 .Ve
1897 .PP
1898 Notes
1899 .IP 1. 5
1900 Notice that it was necessary to define the variable \f(CW\*(C`ax\*(C'\fR. This is
1901 because the \f(CW\*(C`ST\*(C'\fR macro expects it to exist. If we were in an XSUB it
1902 would not be necessary to define \f(CW\*(C`ax\*(C'\fR as it is already defined for
1903 us.
1904 .IP 2. 5
1905 The code
1906 .Sp
1907 .Vb 3
1908 \& SPAGAIN;
1909 \& SP \-= count;
1910 \& ax = (SP \- PL_stack_base) + 1;
1911 .Ve
1912 .Sp
1913 sets the stack up so that we can use the \f(CW\*(C`ST\*(C'\fR macro.
1914 .IP 3. 5
1915 Unlike the original coding of this example, the returned
1916 values are not accessed in reverse order. So \f(CWST(0)\fR refers to the
1917 first value returned by the Perl subroutine and \f(CWST(count\-1)\fR
1918 refers to the last.
1919 .SS "Creating and Calling an Anonymous Subroutine in C"
1920 .IX Subsection "Creating and Calling an Anonymous Subroutine in C"
1921 As we've already shown, \f(CW\*(C`call_sv\*(C'\fR can be used to invoke an
1922 anonymous subroutine. However, our example showed a Perl script
1923 invoking an XSUB to perform this operation. Let's see how it can be
1924 done inside our C code:
1925 .PP
1926 .Vb 1
1927 \& ...
1928 \&
1929 \& SV *cvrv
1930 \& = eval_pv("sub {
1931 \& print \*(AqYou will not find me cluttering any namespace!\*(Aq
1932 \& }", TRUE);
1933 \&
1934 \& ...
1935 \&
1936 \& call_sv(cvrv, G_VOID|G_NOARGS);
1937 .Ve
1938 .PP
1939 \&\f(CW\*(C`eval_pv\*(C'\fR is used to compile the anonymous subroutine, which
1940 will be the return value as well (read more about \f(CW\*(C`eval_pv\*(C'\fR in
1941 "eval_pv" in perlapi). Once this code reference is in hand, it
1942 can be mixed in with all the previous examples we've shown.
1943 .SH "LIGHTWEIGHT CALLBACKS"
1944 .IX Header "LIGHTWEIGHT CALLBACKS"
1945 Sometimes you need to invoke the same subroutine repeatedly.
1946 This usually happens with a function that acts on a list of
1947 values, such as Perl's built-in \fBsort()\fR. You can pass a
1948 comparison function to \fBsort()\fR, which will then be invoked
1949 for every pair of values that needs to be compared. The \fBfirst()\fR
1950 and \fBreduce()\fR functions from List::Util follow a similar
1951 pattern.
1952 .PP
1953 In this case it is possible to speed up the routine (often
1954 quite substantially) by using the lightweight callback API.
1955 The idea is that the calling context only needs to be
1956 created and destroyed once, and the sub can be called
1957 arbitrarily many times in between.
1958 .PP
1959 It is usual to pass parameters using global variables (typically
1960 \&\f(CW$_\fR for one parameter, or \f(CW$a\fR and \f(CW$b\fR for two parameters) rather
1961 than via \f(CW@_\fR. (It is possible to use the \f(CW@_\fR mechanism if you know
1962 what you're doing, though there is as yet no supported API for
1963 it. It's also inherently slower.)
1964 .PP
1965 The pattern of macro calls is like this:
1966 .PP
1967 .Vb 3
1968 \& dMULTICALL; /* Declare local variables */
1969 \& U8 gimme = G_SCALAR; /* context of the call: G_SCALAR,
1970 \& * G_LIST, or G_VOID */
1971 \&
1972 \& PUSH_MULTICALL(cv); /* Set up the context for calling cv,
1973 \& and set local vars appropriately */
1974 \&
1975 \& /* loop */ {
1976 \& /* set the value(s) af your parameter variables */
1977 \& MULTICALL; /* Make the actual call */
1978 \& } /* end of loop */
1979 \&
1980 \& POP_MULTICALL; /* Tear down the calling context */
1981 .Ve
1982 .PP
1983 For some concrete examples, see the implementation of the
1984 \&\fBfirst()\fR and \fBreduce()\fR functions of List::Util 1.18. There you
1985 will also find a header file that emulates the multicall API
1986 on older versions of perl.
1987 .SH "SEE ALSO"
1988 .IX Header "SEE ALSO"
1989 perlxs, perlguts, perlembed
1990 .SH AUTHOR
1991 .IX Header "AUTHOR"
1992 Paul Marquess
1993 .PP
1994 Special thanks to the following people who assisted in the creation of
1995 the document.
1996 .PP
1997 Jeff Okamoto, Tim Bunce, Nick Gianniotis, Steve Kelem, Gurusamy Sarathy
1998 and Larry Wall.
1999 .SH DATE
2000 .IX Header "DATE"
2001 Last updated for perl 5.23.1.